cancel
Showing results for 
Search instead for 
Did you mean: 

How many Diff. Inc's before Full is enforced?

Rhugga23525
Level 3

 

I know with NDMP policies you can only do 9 differentials before a full gets enforced.

 

Does this apply to vanilla networ-based linux client backups done via the netbackup client? I want to do monthly fulls of some workstations and then daily diff's in between.

Thx for any info,

5 REPLIES 5

Andy_Welburn
Level 6

gets enforced"

Really? Is that an "filerism" or a "NBism"? I know there are a limited number of "incremental levels" 0-9 or whatever like the *NIX ufsdump (***EDIT** but again you can follow a 0 with as many 1s 2s or 3s etc as you want) but wasn't aware this carried over to NetBackup.

As far as I was aware you can do as many inc's as you want - obviously for them to have any real use in a DR situation the Full on which the incs are based will still have to be around.

***EDIT #2***

Most of our "vanilla" policies are one FULL per month followed by daily differentials incrementals interspersed with weekly cumulative incrementals, until the next FULL.

Rhugga23525
Level 3

 

I had been told by a symantec engineer at one point over the years that if you configure a monthly full for an NDMP policy and then daily differential's every day in between that the 10th differential will be "upgraded" to a full. (much the same way if you run a differential and there are no current full's in the catalog - a full gets performed) From what I remember it had something to do with the NDMP protocol and not a netbackup limitation.

I've never empirically tested this however. Maybe might be a good time to test this.

 

Andy_Welburn
Level 6

Strange that now you mention it, it's sort of ringing bells! If it was something I was also told then I forgot it almost immediately!

Maybe it does directly link to the dump levels 0-9 mooted earlier, afterall NB just requests that the filer does the "dump" as opposed to a NB client. I must admit tho' that we've never implemented NDMP incrementals so maybe that's why my recollection of this is a little vague!

Will_Restore
Level 6

this describes the process behind incremental backups:

 

NDMP backup levels

At the start of a debug log, you may see an entry titled LEVEL. This entry refers

to an environment variable that NetBackup set based on the type of backup. Here

is an example from a bptm log:

08:48:38.816 [22923] <2> write_data_ndmp: backup environment

values:

08:48:38.816 [22923] <2> write_data_ndmp: Environment 1:

TYPE=dump

08:48:38.816 [22923] <2> write_data_ndmp: Environment 2:

FILESYSTEM=/vol/vol0/2million

08:48:38.817 [22923] <2> write_data_ndmp: Environment 3:

PREFIX=/vol/vol0/2million

08:48:38.817 [22923] <2> write_data_ndmp: Environment 4: LEVEL=0

 

The NDMP backup level is modeled after UNIX dump levels. The backup level is

a number in the range of 0 to 9.

An NDMP backup level of 0 is a full backup. A backup level greater than 0 is an

incremental backup of all objects that were modified since the last backup of a

lower level. For example, level 1 is a backup of all objects that were modified since

the full backup (level 0). Level 3 is a backup of all objects that were modified since

the last level 2 incremental.

 

Table 9-1 NetBackup backup types and corresponding NDMP backup levels

NetBackup backup types NDMP backup levels

NDMP level (last level + 1, up to 9)

never goes higher than 9

Andy_Welburn
Level 6

but doesn't preclude, nor explain, this max number of incrementals......

So, a 0 is full & a 1 anything since the last 0, etc essentially a differential incremental.

Presumably then a subsequent 1 would also be since the last 0 i.e. a cumulative incremental & not since the last 1 i.e. another differential? It's been a while since I last used ufsdump also, so is that how that worked? I remember that being a 1 was anything since the last 0 or last 1........not sure any more!

So it could be,depending on the argument above, that to get a "true" differential it would have to be 0,1,2,3,4,5,6,7,8,& finally 9 which would, therefore, mean that you are limited to 9 "true" differential incrementals, or is my theorising off the mark?